Skip to content

Alexis Roldan · Lancaster, CA

20 Years Building Production AI/ML for Biopharma

Full-Stack Data Scientist & Sr. Data Engineer

I build manufacturing analytics and GenAI systems end to end for Takeda's Los Angeles plasma fractionation plant — the ETL, the modeling, the app, the RAG layer, the deployment. The rule across all of them is the same: code computes the numbers, the model only explains them. Currently at Takeda Pharmaceutical.

Get in touch Download résumé

Lancaster, CA

01who I am

About

Alexis Roldan

I turn biopharma manufacturing data into production systems.

I'm a full-stack data scientist and software developer working on biopharma manufacturing data — twenty years of it, most of that turning plant-floor signals into systems people run a shift on. My biggest deliverables are end to end: I own the ETL, the modeling, the app, the LLM/RAG layer, the deployment and the executive comms. One rule holds across all of them — code computes the numbers, the model only explains them — so every answer traces back to something that was actually measured.

  • End-to-end Data Science & ML
  • Data Engineering & Pipelines
  • Business Intelligence & Visualization
  • Process Optimization & Digital Twins
  • Generative AI & LLM Applications
Lean Six Sigma Green BeltAGILE ChampionDatabricks ChampionMulesoft Champion

English & Spanish (Native) · Japanese (Conversational)

20
Years in biopharma
30+
Solutions shipped
~150
Users on the flagship app
70%
Token reduction on LLM calls

Recognition

Race4Value — FTE Optimization Winner

2023

Star of the Quarter, Q3 FY22

2022

Employee of the Quarter

2017

02what I've built

Selected Work

AMIRA AI manufacturing analytics platform dashboard
Case study

AMIRA

R / Shiny / LLM

Enterprise analytics platform for plasma fractionation — 59% deviation reduction and 70% token optimization via hybrid RAG.

R ShinyDatabricks (Claude Sonnet 4)OSI PIHybrid RAGAWS S3
Read case study
9Deterministic rules
Case study

BioTrack

R / Shiny / RAG

Unifies process, environmental-monitoring, lab, ERP and quality-system data into per-lot investigation records — a nine-rule deterministic engine proposes causes and the LLM may only restate rules that already fired.

R ShinyDatabricks Foundation ModelsDatabricks Vector SearchHybrid RAGJD Edwards (ERP)Plotly
Read case study
Mission Control unified status dashboard
Case study

Mission Control

Next.js / 3D / DevOps

Unifies monitoring across Posit Connect, shinyapps.io and Vercel into one glanceable view — spot a failure in ~2 seconds, reach its logs in one click.

Next.js 15React 19TypeScriptReact Three FiberTanStack QueryAuth.js v5
Read case study
Shiny App Valuation Toolkit dashboard
Case study

Shiny App Valuation Toolkit

R / Shiny

Modified COCOMO II engine for Shiny and data-science repos — ~180 test assertions lock the math, and a v2.0 model audit eliminated cost cliffs of up to 42%.

RShinybslibCOCOMO IIPlotlyQuarto + Typst
Read case study
MIA multimodal intelligent assistant interface
Case study

MIA

Next.js / GenAI

Bilingual learning assistant across 10 personas — streaming OpenAI chat, voice in/out, web search via function calling, document extraction.

Next.js 16React 19TypeScriptOpenAI (streaming + function calling)Google Custom SearchWeb Speech API
Read case study
Aether Wardrobe AI stylist interface
Case study

Aether Wardrobe

Next.js / GenAI

Photograph your wardrobe once; GPT-4o vision catalogs each piece and a deterministic combination engine pairs outfits — the AI only ranks and explains, never invents.

Next.js 16TypeScriptTailwind v4shadcn/uiSupabaseOpenAI GPT-4o (vision)
Read case study
PennyTrail expense tracker dashboard
Case study

PennyTrail

Next.js / TypeScript

Mobile-first PWA for married-filing-jointly tax tracking — GPT-4o spending chat, receipt capture, no database (all state in S3 JSON).

Next.js 16TypeScriptTailwind v4shadcn/uiRechartsAWS S3
Read case study
93%Search-time reduction
Case study

DevSight

R / Shiny / RAG

Semantic search over four years of Trackwise deviations — 93% faster than manual keyword search. Retired in 2026; its retrieval engine still runs inside BioTrack.

R ShinyDatabricks Vector SearchellmerHybrid RAG
Read case study
Horizon CAPA quality management dashboard
Case study

Horizon CAPA Dashboard

R / Shiny / Text-to-SQL

Plain-English querying of CAPAs, Tasks and Deviations — on-demand tool calling keeps prompts at ~500 tokens vs 8–12K. Retired in 2026 in a global application consolidation.

R ShinyquerychatDatabricks GPT-5AWS S3
Read case study
3-phaseSLA logic

LabCast

R / Shiny

QC lab sample tracking with TV/kiosk mode and a forecast model — 3-phase SLA logic across 25 sample types, live state in S3.

R Shinybs4DashForecast modelEDB / LIMS
Internal enterprise app
StreamingLive AI output

Musical Reel Studio

Next.js / GenAI

Generates short-form reel concepts — hook → scene-by-scene → caption → CTA — streamed live, field-by-field from a song, vibe or lyric.

Next.js 15TypeScriptPrismaClerk
Live
20+LLM providers

ReflectAI

React / Express / GenAI

Drafts yearly goals and self-assessments through conversation — provider-agnostic across 20+ LLMs with a mock fallback, exporting polished DOCX/PDF.

React + ViteExpressPluggable LLMsOpenAI / Anthropic / Ollama
Private project — walkthrough on request
RIOT OCR application interface

RIOT — Image OCR App

R / Shiny / AWS

Interactive R/Shiny app extracting text from images via AWS Textract OCR for downstream analytics.

RShinyAWS TextractOCR
Live
TubeScout video discovery dashboard

TubeScout

Python

Finds high-signal YouTube videos using view-to-subscriber ratio analysis — discovery without the algorithm.

PythonStreamlitYouTube API
Repo
YouTube feed dashboard

YouTube Feed Dashboard

R / Shiny / AWS

Personal dashboard surfacing latest videos from favorite channels — AWS-hosted data pipeline, deployed on Posit Connect.

RShinyAWSYouTube API
Live
Voice Assistant AI Streamlit interface

Voice Assistant AI

Python

Python voice assistant with speech recognition, real-time financial data and weather integration.

PythonStreamlitSpeech Recognition
Repo
roldanpack R package

roldanpack

R

Reusable analytics functions and custom visualization themes, installable via GitHub.

RPackage
Repo
0320 years in biopharma

Experience

Baxter 2005–2015Baxalta 2015–2016Shire 2016–2019Takeda 2019–Present

  1. Sr. Data Engineer & Data Scientist (Sr. Manager)

    Apr 2025 – Present

    Takeda Pharmaceutical

    Support the Los Angeles Digital Roadmap, AI/ML development and advanced analytics to drive data-driven decisions.

    • Leading the LA Digital Roadmap and AI/ML development
    • 30+ solutions shipped to production; I own the ETL, the modeling, the app and the deployment
    • Designing, building and maintaining local data pipelines for critical business data
    • Architected hybrid-RAG systems achieving 70% token reduction and 59% deviation reduction
    • Own BioTrack, the site's surviving quality GenAI platform — deterministic hypothesis engine, ERP maintenance evidence, live quality-system integration
    • Carried the retrieval engine out of two consolidated applications into BioTrack, so the capability outlived the apps
    • House rule across every app: code computes the statistics, the LLM only interprets them

  2. Data Scientist III (Sr. Manager)

    Nov 2022 – Apr 2025

    Takeda Pharmaceutical

    Drove the AGILE 4.0 Digital Workstream and cutting-edge analytics to accelerate Takeda's digital capabilities across GMS sites.

    • Drove the AGILE 4.0 Digital Workstream across all GMS sites
    • Executed local and global GMS/GQ big-data and analytics strategy
    • Built and maintained data pipelines and BI applications across all GMS plasma sites

  3. Data Scientist II (Manager)

    Apr 2021 – Nov 2022

    Takeda Pharmaceutical

    Built AI/ML and advanced analytics programs to gain process insight and accelerate digital capabilities.

    • Achieved ~50% reduction in process variation via digital-twin models (PIMS project)
    • Designed A/B tests and developed end-to-end ML applications
    • Los Angeles lead data scientist for global analytics initiatives
04the toolkit

Skills & Education

Languages

RPythonSQLTypeScriptJavaScriptHTML / CSS.NETJava

Data & ML

ShinyDashPlotlyMinitabJMPSIMCA / OPLSApache SparkBase-rate analysisMultiple-comparison correction

Web & Frontend

ReactNext.jsTailwind CSSNode.jsPrismashadcn/ui

Databases

PostgreSQLSQL ServerOracleSupabaseNeon

Cloud & DevOps

AWSDatabricksVercelDockerGitPosit Connect

BI Tools

Qlik SensePower BITableau

AI & GenAI

LLM IntegrationRAG ArchitectureVector SearchTool CallingPrompt EngineeringText-to-SQLAgentic AIClaude / GPT APIsVercel AI SDKVoice AI (TTS/STT)

Methodologies

Six SigmaSCRUM / AgileDMAICMLOpsProject Management

Education & Certifications

  • 2026

    Build Data Pipelines with Lakeflow Spark Declarative Pipelines — Databricks Academy

  • 2026

    Data Warehousing with Databricks — Databricks Academy

  • 2026

    Building Single-Agent Applications on Databricks — Databricks Academy

  • 2023

    Oxford Machine Learning Summer School — AI for Global Goals

  • 2023

    ODSC AI Bootcamp — Open Data Science Conference

  • 2020

    Data Science Professional — HarvardX

  • 2020

    Statistical Learning, Inference & Modeling — HarvardX / Stanford Online

  • 2012

    Computer Science — Cal State Northridge

05let's talk

Get in touch

Have a project or role in mind?

I'm always open to conversations about data engineering, applied ML, and GenAI in regulated/manufacturing environments. Drop a note and I'll get back to you.